The DeLorean Parallel: Lessons from Automotive Innovation

Krishnan opened her presentation with an unexpected analogy—the iconic DeLorean sports car from the Back to the Future films. The story of John DeLorean, a rising star in the automotive industry, offers striking parallels to today's AI landscape.

John DeLorean recognised unethical practices within the automotive industry and envisioned creating a car that was safe, fuel-efficient, and affordable. His forward-thinking approach led him to incorporate features that weren't legally required at the time:

• Large rear taillights for enhanced night-time visibility
• Airbags (planned but delayed due to production challenges)
• Third brake lights for additional safety 

These innovations were considered visionary but not necessary. Today, however, they're mandatory features in every vehicle—a testament to how industry standards evolve from optional best practices to regulatory requirements.

"I'm telling you this story because I just want you to think about a parallel in the AI industry as well," Krishnan explained. "Just like the automotive industry, there is a shift that has happened in the AI industry, where there is a clear before and after picture."

The Four Key Trends Shaping AI Governance

Krishnan identified four critical trends emerging from IBM's work with clients across industries. These challenges represent the primary obstacles preventing organisations from realising AI's full potential.

1. Operationalising AI with Confidence

The journey from AI experimentation to production deployment remains fraught with challenges. Krishnan shared a striking example: one client had built 700 AI models but had no visibility into how they were constructed, what stage of development they were in, or how to monitor them effectively.

Key challenges include:

• Fragmented landscapes where models are built using disparate tools without centralised oversight.
• Lack of visibility preventing informed decision-making and production deployment.
• Insufficient transparency and explainability throughout the entire AI lifecycle, not just in production.

"More often than not, when we think about model explainability and transparency, we think about models that are already in production. But that's not the case," Krishnan emphasised. "You have to think entire lifecycle. You have to think about even before something gets built: am I using the right data for this? Is this the right kind of model? Do I have bias in my data?"

The automation imperative:

Without automation, scaling AI initiatives becomes impossible. Manual monitoring and tracking might work for a handful of models, but as organisations develop more applications and process increasing volumes of data, automated governance becomes essential.

2. Managing Risk and Reputation Through Responsible AI

Consumer trust has become paramount in the AI era. Organisations that demonstrate ethical AI practices earn customer loyalty, whilst those that fail face lasting reputational damage.

"Once the trust is lost, it's really hard to get it back. Nobody wants to be in the press for the wrong reasons," Krishnan noted.

Progressive organisations are moving beyond reactive risk management to proactive ethical AI strategies. Rather than waiting for problems to emerge, leading companies are embedding ethical and transparent AI principles into their strategic imperatives from the outset.

"We don't want to catch it after the shoe drops, right? We want to be able to proactively think about this as a first principle of design," Krishnan advised.

3. Navigating the Evolving Regulatory Landscape

Similar to how automotive safety features transitioned from optional innovations to legal requirements, AI regulations are rapidly proliferating across industries and jurisdictions. Organisations must prepare for compliance with an expanding array of AI-specific regulations whilst maintaining operational agility.

4. Engaging Diverse Stakeholders

The AI governance playing field has fundamentally changed. No longer confined to data science teams, AI initiatives now involve stakeholders across the entire organisation—from legal and compliance to operations and executive leadership.

"Everybody is involved, and everybody has a stake to make AI successful in the business," Krishnan observed.

Key Takeaways for Organisations

Krishnan's keynote addressed three fundamental questions facing organisations today:

1. How does my organisation operationalise AI with confidence? 
2. How do we better manage AI risk to avoid brand degradation? 
3. How does my organisation scale whilst complying with growing AI regulations? 

The path forward requires comprehensive, open, and automated AI governance solutions that provide end-to-end lifecycle visibility, embed ethical principles from design through deployment, and enable organisations to scale responsibly.

As AI continues its rapid evolution from experimental technology to business-critical infrastructure, governance frameworks will transition from competitive advantage to baseline requirement—much like those safety features in the DeLorean that once seemed visionary but are now simply expected.

The Acceleration of AI Regulations

The regulatory landscape represents a powerful external force reshaping how organisations approach AI governance. Krishnan drew parallels to the evolution of data governance, which transitioned from voluntary principles and strategies to mandatory regulations—a shift AI is now experiencing at unprecedented speed.

"AI is going through the same shift here. These were strategies in the past, but now these are actually getting translated into real policies that companies have to follow," Krishnan explained.

The pace of regulatory change:

• Historical pattern: New regulations emerged once every year or two 
• Current reality: Two or more regulations annually—and accelerating 
• Global reach: Regulations appearing across all industries and jurisdictions

Recent examples include New York City's hiring law and new United States Government regulations for AI-based recruitment—demonstrating how quickly theoretical concerns are becoming legal requirements.

"It's coming rapidly. It's coming across the globe, and it's coming to every industry. It's not just secluded in one set of industries here," Krishnan noted. The challenge for organisations extends beyond current compliance to building frameworks that can adapt to tomorrow's regulatory environment.

The financial stakes of non-compliance:

The consequences of regulatory violations extend far beyond reputational damage. The EU AI Act, for instance, can impose fines of up to 6% of a company's global revenue—a penalty substantial enough to impact any organisation's bottom line.

The Expanding Stakeholder Ecosystem

IBM's philosophy that "data science is a team sport" has evolved to encompass an even broader playing field. AI governance now requires coordination amongst stakeholders far beyond data science and model validation teams.

Key stakeholders now involved in AI governance:

• Chief Financial Officers (CFOs): Managing profitability risks and potential regulatory fines 
• Chief Marketing Officers (CMOs): Protecting brand reputation and customer trust 
• Legal and compliance teams: Ensuring regulatory adherence 
• Data scientists and engineers: Building and deploying models 
• Model validators: Ensuring quality and fairness
• Executive leadership: Setting strategic direction

"There's multiple stakeholders in an enterprise that are actually coming together to make sure that your AI is ethical, it's governed, it's effective, and it gives what it's supposed to do," Krishnan emphasised.

Reframing AI Governance: The Formula One Analogy

Krishnan challenged the common perception that governance slows innovation, offering a compelling analogy to reframe the conversation.

"Think of even a Formula One car, for instance, right? The idea of these cars, there are safety checks inside the car, there are so many people putting up so many things to make sure that the car is safe. But none of that is to slow down the car, it's actually to help the car get faster, safely."

This perspective shift is crucial. AI governance isn't designed to impede progress, it's engineered to accelerate value realisation whilst maintaining appropriate safeguards.

"Very often people think about the word of governance and associate it negatively. But that's really not the case. When you think about AI governance, it's actually designed to help you get value from AI faster, with guardrails around it."

Three Pillars of Effective AI Governance Solutions

Based on extensive client work, IBM has identified three essential capabilities that form the foundation of robust AI governance.

1. Lifecycle Governance

Comprehensive monitoring, cataloguing, and understanding of models throughout their entire lifecycle, from data selection through production deployment.

Real-world impact:

One client's data science team spent one to two months building models, which then required another two months for a small validation team to review. The validation process involved constant back-and-forth questions about data sources, model variations, and methodology, often communicated through Excel spreadsheets.

The challenges compounded when:

• Data scientists had forgotten details from months earlier.
• Team members had left the organisation.
• No automated documentation existed.

By implementing lifecycle governance, IBM reduced this timeline from months to weeks, eliminating manual documentation and creating visibility for all stakeholders throughout the model development process.

2. Risk Management Through Customised Workflows

Creating tailored dashboards and automated workflows that provide relevant information to each stakeholder whilst maintaining consistent oversight.

Case study: Cross-regional model reuse:

A data science team selected an existing production model for a similar use case, aiming to accelerate development. However, they discovered the original model was built for United States customers with specific business controls, whilst the new model targeted EU customers with entirely different regulatory requirements.

Without proper workflow visibility and business controls, teams waste time building inappropriate solutions, leading to validation failures and repeated rework cycles. Automated workflows with clear business controls prevent such misalignments from the outset.

3. Regulatory Compliance Translation

Converting complex regulations into actionable business controls that integrate seamlessly into organisational workflows.

Client perspectives:

Organisations increasingly request that IBM translate emerging regulations into practical implementation guidelines, removing this burden from data science and validation teams. This allows technical teams to focus on model development whilst ensuring automatic compliance.

The hidden bias challenge:

One client committed to fair hiring practices removed obvious attributes like gender from their recruitment models. However, the model included "willingness to work night shifts"—an attribute that introduced indirect gender bias when women self-selected out of night shift roles.

"You need tools that will be able to catch those kinds of indirect biases as well. If you did not do that, again, you're going to be out of compliance with regulations," Krishnan warned.

Essential Characteristics of AI Governance Solutions

Beyond the three core capabilities, Krishnan identified three critical attributes that distinguish effective governance solutions.

Comprehensive Coverage

Governance cannot be limited to production monitoring. Effective solutions must orchestrate oversight throughout the entire AI lifecycle:

• Data acquisition: Ensuring data is free from bias at the source 
• Model development: Monitoring experimentation and methodology 
• Testing and validation: Verifying performance and fairness 
• Production deployment: Ongoing monitoring and maintenance 

"I cannot stress this enough, because it is not enough that you have put in some model metrics and monitoring in your production and then you're happy to go," Krishnan emphasised.

Open Architecture

Organisations rarely build all models using a single technology stack. Data scientists prefer tools they're comfortable with, resulting in heterogeneous environments.

Effective governance solutions must augment existing technologies and processes rather than requiring wholesale replacement. The client with 700 models, for instance, had built them using various tools—any governance solution needed to work across this diverse landscape.

Automation at Scale

Manual governance processes might work for a handful of models but become impossible as organisations scale their AI initiatives.

"None of this would work until it is automated, so that you can operationalise it at scale," Krishnan stated definitively.

Beyond Technology: The People-Process-Technology Trifecta

Krishnan concluded by emphasising that technology alone cannot deliver successful AI governance.

"A good AI governance solution has the trifecta of people, process and technology together."

IBM's approach begins with workshops that address fundamental questions:

• Who are the stakeholders invested in AI success?
• Can all necessary parties commit to the initiative?
• How has the stakeholder map expanded beyond traditional roles? 

This holistic approach, combining the right people, well-designed processes, and appropriate technology, creates the foundation for AI governance that truly accelerates business value whilst maintaining ethical standards and regulatory compliance.

About the Speaker:

Priya Krishnan leads product management for data science and AI governance within IBM's Data and AI portfolio, where she focuses on helping organisations deploy AI responsibly and effectively.